imhcode 1.0.0

package/skills/development-workflows/rpi/.claude/commands/rpi/research.md

@@ -0,0 +1,381 @@
+---
+description: Research and analyze feature viability - GO/NO-GO decision gate
+argument-hint: "<feature-slug>"
+---
+
+## User Input
+
+```text
+$ARGUMENTS
+```
+
+You **MUST** parse the user input to extract the feature slug (the folder name in `rpi/`).
+
+**Expected Input Format**: `rpi/{feature-slug}/REQUEST.md`
+
+## Purpose
+
+This command performs comprehensive research and analysis of feature requests **before** the planning phase begins. It acts as a critical GO/NO-GO gate to determine whether a feature idea should proceed to detailed planning.
+
+**Key Objectives**:
+- Assess product-market fit and user value
+- Evaluate technical feasibility and complexity
+- Identify risks and potential blockers
+- Determine the right approach (build, buy, partner, or decline)
+- Make go/no-go recommendation with clear rationale
+
+**Prerequisites**:
+- Feature folder exists at `rpi/{feature-slug}/`
+- Feature request file exists at `rpi/{feature-slug}/REQUEST.md`
+
+**Output Location**: `rpi/{feature-slug}/research/RESEARCH.md`
+
+**This is Step 2 of the RPI Workflow** (after initial feature description in Step 1).
+
+## Outline
+
+1. **Load Context**: Read feature description from `rpi/{feature-slug}/` and project constitution (if exists)
+2. **Parse Feature Request**: Use requirement-parser agent to extract structured requirements
+3. **Execute Multi-Phase Research**:
+   - Phase 1: Parse Feature Request (requirement-parser agent)
+   - Phase 2: Product Analysis with Constitution Alignment (product-manager agent)
+   - Phase 2.5: Technical Discovery (Explore agent) - **CRITICAL: Deep code exploration**
+   - Phase 3: Technical Feasibility (senior-software-engineer agent)
+   - Phase 4: Strategic Assessment (technical-cto-advisor agent)
+   - Phase 5: Generate Research Report (documentation-analyst-writer agent)
+4. **Synthesize Recommendation**: Combine all analyses into clear go/no-go recommendation
+5. **Validate Output**: Check against quality gates
+6. **Report Completion**: Provide recommendation, next steps, and report location
+
+## Phases
+
+### Phase 0: Load Context
+
+**Prerequisites**: Feature slug provided, `rpi/{feature-slug}/REQUEST.md` exists
+
+**Process**:
+1. **Read feature description**:
+   - Read `rpi/{feature-slug}/REQUEST.md` (required)
+   - Extract feature requirements and goals from REQUEST.md
+
+2. **Check for project constitution** (optional):
+   - Look for a constitution or principles document in the repository
+   - Common locations: `constitution.md`, `PRINCIPLES.md`, `.project/constitution.md`
+   - If found, extract core principles, constraints, and objectives
+
+3. **Create research context**:
+   - Synthesize into concise summary for agents
+   - Identify key alignment criteria
+
+**Outputs**:
+- Feature description summary
+- Constitutional principles (if found)
+- Alignment criteria for evaluation
+
+**Validation**:
+- [ ] Feature folder exists in `rpi/{feature-slug}/`
+- [ ] Feature description extracted
+- [ ] Constitution checked and loaded (if exists)
+
+---
+
+### Phase 1: Parse Feature Request
+
+**Prerequisites**: Phase 0 complete
+
+**Agent**: requirement-parser (planning domain)
+
+**Process**:
+1. **Launch requirement-parser agent** with feature description
+2. **Agent extracts**:
+   - Feature name and type
+   - Target component(s)
+   - Goals and objectives
+   - Functional and non-functional requirements
+   - Constraints and assumptions
+   - Complexity estimate
+   - Clarifying questions (if any)
+
+3. **Review parsing results**:
+   - If clarifying questions exist, **STOP and ask user** before proceeding
+
+**Outputs**:
+- Structured requirements document
+- Feature metadata (name, type, component, complexity)
+- Clarifying questions (if any)
+
+---
+
+### Phase 2: Product Analysis with Constitution Alignment
+
+**Prerequisites**: Phase 1 complete, requirements clear
+
+**Agent**: product-manager
+
+**Process**:
+1. **Launch product-manager agent** with:
+   - Parsed requirements from Phase 1
+   - Constitutional context from Phase 0
+
+2. **Agent analyzes**:
+   - **User Value**: Who benefits? How much impact?
+   - **Market Fit**: Does this align with market needs?
+   - **Product Vision**: Does this fit our product strategy?
+   - **Constitutional Alignment**: Does this align with project principles?
+   - **Constraints Check**: Does this violate any constitutional constraints?
+
+3. **Agent provides**:
+   - Product viability score (High/Medium/Low)
+   - User value assessment
+   - Strategic alignment evaluation
+   - Priority recommendation
+   - Product concerns or red flags
+
+**Outputs**:
+- Product viability assessment
+- User value analysis
+- Strategic alignment score
+- Constitutional alignment summary (if applicable)
+
+---
+
+### Phase 2.5: Technical Discovery (Code Exploration)
+
+**Prerequisites**: Phases 1-2 complete, product viability established
+
+**Agent**: Explore (via Task tool with subagent_type="Explore")
+
+**Purpose**: **CRITICAL PHASE** - Deeply analyze existing codebase BEFORE making technical feasibility assessment.
+
+**Process**:
+1. **Launch Explore agent** with target component(s)
+2. **Agent investigates**:
+   - **Existing Implementation**: What code already exists for similar functionality?
+   - **Integration Points**: What systems/modules would this feature touch?
+   - **Current Architecture**: How is the current system structured?
+   - **Data Models**: What database schemas or data structures exist?
+   - **Dependencies**: What libraries, services are already integrated?
+   - **Existing Patterns**: What coding patterns and conventions are used?
+
+3. **Agent provides**:
+   - **Current State Summary**: What exists today
+   - **Integration Analysis**: Where proposed feature would fit
+   - **Code Conflicts**: What would break or conflict
+   - **Leverage Opportunities**: What can be reused vs rebuilt
+   - **Technical Constraints**: Real constraints from existing code
+
+**Outputs**:
+- Current implementation summary
+- Integration points map
+- Code conflicts identified
+- Reusable components identified
+- Technical constraints from code
+
+**Critical**: This phase ensures Phase 3 is based on **actual code reality**, not assumptions.
+
+---
+
+### Phase 3: Technical Feasibility Assessment
+
+**Prerequisites**: Phases 1-2.5 complete, code explored
+
+**Agent**: senior-software-engineer
+
+**Process**:
+1. **Launch senior-software-engineer agent** with:
+   - Parsed requirements from Phase 1
+   - Product context from Phase 2
+   - **Technical discovery results from Phase 2.5**
+
+2. **Agent analyzes** (informed by Phase 2.5 discoveries):
+   - **Technical Approach**: What are the implementation options?
+   - **Complexity**: How difficult is this to build?
+   - **Dependencies**: What systems/services are needed?
+   - **Technical Debt**: Will this create or reduce tech debt?
+   - **Risks**: What are the technical risks?
+
+3. **Agent provides**:
+   - Technical feasibility score (High/Medium/Low)
+   - Recommended approach (with alternatives)
+   - Complexity estimate (Simple/Medium/Complex)
+   - Technical risks and mitigations
+
+**Outputs**:
+- Technical feasibility score
+- Recommended implementation approach
+- Complexity and effort estimate
+- Technical risks and mitigations
+
+---
+
+### Phase 4: Strategic Assessment
+
+**Prerequisites**: Phases 1-3 complete
+
+**Agent**: technical-cto-advisor
+
+**Process**:
+1. **Launch technical-cto-advisor agent** with all previous phase outputs
+
+2. **Agent synthesizes**:
+   - **Overall Assessment**: Combine product + technical perspectives
+   - **Strategic Alignment**: Does this align with engineering principles AND project constitution?
+   - **Risk vs. Reward**: Is the value worth the effort and risk?
+   - **Alternative Options**: Build, buy, partner, defer, or decline?
+
+3. **Agent provides**:
+   - **Go/No-Go Recommendation**: Clear decision with confidence level
+   - **Rationale**: Detailed reasoning
+   - **Recommended Approach**: If "go", what's the best path forward?
+   - **Conditions**: Any prerequisites for proceeding?
+   - **Risks**: Key risks if we proceed
+
+**Outputs**:
+- Go/No-Go recommendation
+- Strategic rationale
+- Recommended approach
+- Risk summary
+
+---
+
+### Phase 5: Generate Research Report
+
+**Prerequisites**: Phases 1-4 complete
+
+**Agent**: documentation-analyst-writer (via Task tool)
+
+**Process**:
+1. **Launch documentation-analyst-writer agent** with all phase outputs
+
+2. **Agent generates report** with sections:
+   - **Executive Summary**: One-paragraph overview with recommendation
+   - **Feature Overview**: Name, type, component, goals
+   - **Requirements Summary**: Key functional and non-functional requirements
+   - **Product Analysis**: User value, market fit, strategic alignment
+   - **Technical Discovery**: Current state, integration points, constraints from code
+   - **Technical Analysis**: Feasibility, approach, complexity, risks
+   - **Strategic Recommendation**: Go/no-go with detailed rationale
+   - **Next Steps**: What to do based on recommendation
+
+3. **Agent creates markdown file**: `rpi/{feature-slug}/research/RESEARCH.md`
+
+**Outputs**:
+- Complete research report saved to `rpi/{feature-slug}/research/RESEARCH.md`
+
+---
+
+## Sub-Agent Delegation
+
+This command orchestrates 6 specialist agents:
+
+| Phase | Agent | Type | Location |
+|-------|-------|------|----------|
+| Phase 1 | requirement-parser | Custom | .claude/agents/requirement-parser.md |
+| Phase 2 | product-manager | Custom | .claude/agents/product-manager.md |
+| Phase 2.5 | Explore | Built-in | Task tool with subagent_type="Explore" |
+| Phase 3 | senior-software-engineer | Custom | .claude/agents/senior-software-engineer.md |
+| Phase 4 | technical-cto-advisor | Custom | .claude/agents/technical-cto-advisor.md |
+| Phase 5 | documentation-analyst-writer | Built-in | Task tool with subagent_type="documentation-analyst-writer" |
+
+---
+
+## Completion Report
+
+Report the following on successful completion:
+
+### Research Recommendation
+
+**Decision**: [GO | NO-GO | CONDITIONAL GO | DEFER]
+
+**Confidence**: [High | Medium | Low]
+
+**Rationale** (1-2 sentences):
+[Key reasons for recommendation]
+
+---
+
+### Research Summary
+
+**Feature**: {feature-name}
+**Type**: {feature-type}
+**Component**: {target-component}
+**Complexity**: {Simple | Medium | Complex}
+
+**Scores**:
+- Product Viability: [High/Medium/Low]
+- Technical Feasibility: [High/Medium/Low]
+- Overall Assessment: [High/Medium/Low]
+
+**Key Risks**:
+1. {risk-1}
+2. {risk-2}
+3. {risk-3}
+
+---
+
+### Report Location
+
+**Full Research Report**: `rpi/{feature-slug}/research/RESEARCH.md`
+
+---
+
+### Next Steps
+
+Based on the **[GO/NO-GO]** recommendation:
+
+**If GO**:
+1. Review the research report: `rpi/{feature-slug}/research/RESEARCH.md`
+2. Proceed to planning: `/rpi:plan "{feature-slug}"`
+
+**If CONDITIONAL GO**:
+1. Review conditions in report
+2. Address conditions before proceeding
+3. Re-run research if needed
+
+**If DEFER**:
+1. Review timeline recommendation in report
+2. Revisit when timing is appropriate
+
+**If NO-GO**:
+1. Review rationale in report
+2. Consider alternatives mentioned
+3. Archive for future reference
+
+---
+
+## Error Handling
+
+**If REQUEST.md doesn't exist**:
+- Action: Stop and inform user
+- Message: "Feature request file `rpi/{feature-slug}/REQUEST.md` not found. Create the feature folder and REQUEST.md first (Step 1: Describe in Plan Mode)."
+
+**If feature description is too vague**:
+- Action: requirement-parser will identify clarifying questions
+- Message: "Need more information. Please answer:"
+- Next: Wait for answers, then proceed
+
+**If agents fail or timeout**:
+- Action: Retry once
+- Next: If retry fails, ask user whether to continue with incomplete research
+
+---
+
+## Notes
+
+- **When to Use**: After Step 1 (Describe) creates the feature folder
+- **Critical Gate**: This prevents wasted effort on non-viable features
+- **Part of RPI Workflow**: Step 2 of 4 (Describe → Research → Plan → Implement)
+
+---
+
+## Post-Completion Action
+
+**IMPORTANT**: After completing the research workflow, ALWAYS prompt the user to compact the conversation:
+
+> **Context Management**: This research workflow consumed significant context. To free up space for the next steps, please run:
+>
+> ```
+> /compact
+> ```
+>
+> This will summarize the conversation and preserve important findings while reducing token usage for subsequent commands.

package/skills/development-workflows/rpi/rpi-workflow.md

@@ -0,0 +1,101 @@
+# RPI Workflow
+
+**RPI** = **R**esearch → **P**lan → **I**mplement
+
+A systematic development workflow with validation gates at each phase. Prevents wasted effort on non-viable features and ensures comprehensive documentation.
+
+<table width="100%">
+<tr>
+<td><a href="../../">← Back to Claude Code Best Practice</a></td>
+<td align="right"><img src="../../!/claude-jumping.svg" alt="Claude" width="60" /></td>
+</tr>
+</table>
+
+---
+
+## Overview
+
+![RPI Workflow](rpi-workflow.svg)
+
+---
+
+## Installation
+
+Copy the `.claude` folder (containing `agents/` and `commands/rpi/`) to your repository root, then create the `rpi/plans` directory.
+
+---
+
+## Example Workflow
+
+### Feature: User Authentication
+
+**Step 1: Describe**
+```
+User: "Add OAuth2 authentication with Google and GitHub providers"
+
+1. Claude generates plan
+   → Output: rpi/plans/oauth2-authentication.md
+2. Create feature folder: rpi/oauth2-authentication/
+3. Copy the plan into the feature folder
+4. Rename the plan to REQUEST.md
+   → Final: rpi/oauth2-authentication/REQUEST.md
+```
+
+**Step 2: Research**
+```bash
+/rpi:research rpi/oauth2-authentication/REQUEST.md
+```
+Output:
+- `research/RESEARCH.md` with analysis
+- Verdict: **GO** (feasible, aligned with strategy)
+
+**Step 3: Plan**
+```bash
+/rpi:plan oauth2-authentication
+```
+Output:
+- `plan/pm.md` - User stories and acceptance criteria
+- `plan/ux.md` - Login UI flows
+- `plan/eng.md` - Technical architecture
+- `plan/PLAN.md` - 3 phases, 15 tasks
+
+**Step 4: Implement**
+```bash
+/rpi:implement oauth2-authentication
+```
+Progress:
+- Phase 1: Backend Foundation → PASS
+- Phase 2: Frontend Integration → PASS
+- Phase 3: Testing & Polish → PASS
+
+Result: Feature complete, ready for PR.
+
+---
+
+## Feature Folder Structure
+
+All feature work lives in `rpi/{feature-slug}/`:
+
+```
+rpi/{feature-slug}/
+├── REQUEST.md              # Step 1: Initial feature description
+├── research/
+│   └── RESEARCH.md         # Step 2: GO/NO-GO analysis
+├── plan/
+│   ├── PLAN.md             # Step 3: Implementation roadmap
+│   ├── pm.md               # Product requirements
+│   ├── ux.md               # UX design
+│   └── eng.md              # Technical specification
+└── implement/
+    └── IMPLEMENT.md        # Step 4: Implementation record
+```
+
+---
+
+## Agents and Commands
+
+| Command | Agents Used |
+|---------|-------------|
+| `/rpi:research` | requirement-parser, product-manager, Explore, senior-software-engineer, technical-cto-advisor, documentation-analyst-writer |
+| `/rpi:plan` | senior-software-engineer, product-manager, ux-designer, documentation-analyst-writer |
+| `/rpi:implement` | Explore, senior-software-engineer, code-reviewer |

package/skills/development-workflows/rpi/rpi-workflow.svg

@@ -0,0 +1,178 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 900 980">
+  <defs>
+    <linearGradient id="prereqGrad" x1="0%" y1="0%" x2="100%" y2="0%">
+      <stop offset="0%" style="stop-color:#374151;stop-opacity:1" />
+      <stop offset="100%" style="stop-color:#4b5563;stop-opacity:1" />
+    </linearGradient>
+    <linearGradient id="step1Grad" x1="0%" y1="0%" x2="100%" y2="0%">
+      <stop offset="0%" style="stop-color:#1f2937;stop-opacity:1" />
+      <stop offset="100%" style="stop-color:#374151;stop-opacity:1" />
+    </linearGradient>
+    <linearGradient id="step2Grad" x1="0%" y1="0%" x2="100%" y2="0%">
+      <stop offset="0%" style="stop-color:#111827;stop-opacity:1" />
+      <stop offset="100%" style="stop-color:#1f2937;stop-opacity:1" />
+    </linearGradient>
+    <linearGradient id="step3Grad" x1="0%" y1="0%" x2="100%" y2="0%">
+      <stop offset="0%" style="stop-color:#1f2937;stop-opacity:1" />
+      <stop offset="100%" style="stop-color:#374151;stop-opacity:1" />
+    </linearGradient>
+    <linearGradient id="step4Grad" x1="0%" y1="0%" x2="100%" y2="0%">
+      <stop offset="0%" style="stop-color:#374151;stop-opacity:1" />
+      <stop offset="100%" style="stop-color:#4b5563;stop-opacity:1" />
+    </linearGradient>
+    <filter id="shadow" x="-20%" y="-20%" width="140%" height="140%">
+      <feDropShadow dx="2" dy="2" stdDeviation="3" flood-opacity="0.15"/>
+    </filter>
+    <style>
+      .title { font-family: 'Inter', 'Segoe UI', Arial, sans-serif; font-size: 28px; font-weight: 600; fill: #111827; letter-spacing: -0.5px; }
+      .subtitle { font-family: 'Inter', 'Segoe UI', Arial, sans-serif; font-size: 14px; fill: #6b7280; letter-spacing: 2px; text-transform: uppercase; }
+      .box-title { font-family: 'Inter', 'Segoe UI', Arial, sans-serif; font-size: 16px; font-weight: 600; fill: #f9fafb; letter-spacing: 0.5px; }
+      .box-text { font-family: 'Inter', 'Segoe UI', Arial, sans-serif; font-size: 12px; fill: #e5e7eb; }
+      .output-text { font-family: 'JetBrains Mono', 'Courier New', monospace; font-size: 11px; fill: #374151; }
+      .arrow-text { font-family: 'Inter', 'Segoe UI', Arial, sans-serif; font-size: 11px; fill: #9ca3af; font-weight: 500; }
+      .step-number { font-family: 'Inter', 'Segoe UI', Arial, sans-serif; font-size: 11px; font-weight: 600; fill: #9ca3af; }
+      .gate-text { font-family: 'Inter', 'Segoe UI', Arial, sans-serif; font-size: 10px; font-weight: 600; fill: #374151; text-transform: uppercase; letter-spacing: 0.5px; }
+      .agent-text { font-family: 'Inter', 'Segoe UI', Arial, sans-serif; font-size: 10px; fill: #9ca3af; }
+      .clickable { cursor: pointer; transition: all 0.2s ease; }
+      .clickable:hover { filter: brightness(1.15); }
+    </style>
+  </defs>
+
+  <!-- Background -->
+  <rect width="900" height="980" fill="#fafafa"/>
+
+  <!-- Title -->
+  <text x="450" y="40" text-anchor="middle" class="title">RPI Workflow</text>
+  <text x="450" y="60" text-anchor="middle" class="subtitle">Research → Plan → Implement</text>
+
+  <!-- Prerequisites Box: y=80, height=115, ends at 195 -->
+  <g class="clickable" transform="translate(100, 80)">
+    <rect width="700" height="115" rx="8" fill="url(#prereqGrad)" filter="url(#shadow)"/>
+    <rect width="700" height="36" rx="8" fill="rgba(0,0,0,0.2)"/>
+    <rect y="28" width="700" height="8" fill="rgba(0,0,0,0.2)"/>
+    <text x="24" y="24" class="box-title">PREREQUISITES</text>
+    <g transform="translate(24, 50)">
+      <text x="0" y="12" class="step-number">01</text>
+      <text x="28" y="12" class="box-text">Create constitution.md at project root</text>
+    </g>
+    <g transform="translate(24, 72)">
+      <text x="0" y="12" class="step-number">02</text>
+      <text x="28" y="12" class="box-text">Add { "plansDirectory": "./rpi/plans" } to .claude/settings.json</text>
+    </g>
+    <g transform="translate(24, 94)">
+      <text x="0" y="12" class="step-number">03</text>
+      <text x="28" y="12" class="box-text">Create rpi/plans directory</text>
+    </g>
+  </g>
+
+  <!-- Arrow 1: between 195 and 210 -->
+  <g transform="translate(450, 198)">
+    <line x1="0" y1="0" x2="0" y2="10" stroke="#d1d5db" stroke-width="2"/>
+    <polygon points="0,16 -5,8 5,8" fill="#d1d5db"/>
+  </g>
+
+  <!-- Step 1: y=218, height=175, ends at 393 -->
+  <g class="clickable" transform="translate(100, 218)">
+    <rect width="700" height="175" rx="8" fill="url(#step1Grad)" filter="url(#shadow)"/>
+    <rect width="700" height="36" rx="8" fill="rgba(0,0,0,0.3)"/>
+    <rect y="28" width="700" height="8" fill="rgba(0,0,0,0.3)"/>
+    <text x="24" y="24" class="box-title">STEP 1 — DESCRIBE</text>
+    <g transform="translate(24, 50)">
+      <text x="0" y="12" class="step-number">01</text>
+      <text x="28" y="12" class="box-text">Open Claude Code in plan mode and describe your feature</text>
+    </g>
+    <rect x="24" y="70" width="300" height="24" rx="4" fill="#f3f4f6"/>
+    <text x="34" y="86" class="output-text">→ Output: rpi/plans/{plan}.md</text>
+    <g transform="translate(24, 105)">
+      <text x="0" y="12" class="step-number">02</text>
+      <text x="28" y="12" class="box-text">Create a feature folder inside rpi/ with your feature name</text>
+    </g>
+    <g transform="translate(24, 127)">
+      <text x="0" y="12" class="step-number">03</text>
+      <text x="28" y="12" class="box-text">Copy the plan into the feature folder and rename to REQUEST.md</text>
+    </g>
+    <rect x="24" y="148" width="340" height="24" rx="4" fill="#f3f4f6"/>
+    <text x="34" y="164" class="output-text">→ Final: rpi/{feature-slug}/REQUEST.md</text>
+  </g>
+
+  <!-- Arrow 2: between 393 and 408 -->
+  <g transform="translate(450, 396)">
+    <line x1="0" y1="0" x2="0" y2="10" stroke="#d1d5db" stroke-width="2"/>
+    <polygon points="0,16 -5,8 5,8" fill="#d1d5db"/>
+  </g>
+
+  <!-- Step 2: y=416, height=130, ends at 546 -->
+  <g class="clickable" transform="translate(100, 416)">
+    <rect width="700" height="130" rx="8" fill="url(#step2Grad)" filter="url(#shadow)"/>
+    <rect width="700" height="36" rx="8" fill="rgba(255,255,255,0.08)"/>
+    <rect y="28" width="700" height="8" fill="rgba(255,255,255,0.08)"/>
+    <text x="24" y="24" class="box-title">STEP 2 — RESEARCH</text>
+    <rect x="580" y="8" width="100" height="20" rx="10" fill="#fbbf24"/>
+    <text x="630" y="22" text-anchor="middle" class="gate-text">GO / NO-GO</text>
+    <rect x="24" y="48" width="420" height="26" rx="4" fill="rgba(255,255,255,0.1)"/>
+    <text x="34" y="66" class="box-text" style="font-family: 'JetBrains Mono', monospace; font-size: 11px;">/rpi:research rpi/{feature-slug}/REQUEST.md</text>
+    <rect x="24" y="82" width="400" height="24" rx="4" fill="#f3f4f6"/>
+    <text x="34" y="98" class="output-text">→ Output: rpi/{feature-slug}/research/RESEARCH.md</text>
+    <text x="460" y="62" class="agent-text">Agents: requirement-parser, product-manager,</text>
+    <text x="460" y="76" class="agent-text">Explore, senior-software-engineer,</text>
+    <text x="460" y="90" class="agent-text">technical-cto-advisor, documentation-analyst-writer</text>
+  </g>
+
+  <!-- Arrow 3: between 546 and 561 -->
+  <g transform="translate(450, 549)">
+    <line x1="0" y1="0" x2="0" y2="10" stroke="#d1d5db" stroke-width="2"/>
+    <polygon points="0,16 -5,8 5,8" fill="#d1d5db"/>
+    <text x="15" y="12" class="arrow-text">if GO</text>
+  </g>
+
+  <!-- Step 3: y=569, height=120, ends at 689 -->
+  <g class="clickable" transform="translate(100, 569)">
+    <rect width="700" height="120" rx="8" fill="url(#step3Grad)" filter="url(#shadow)"/>
+    <rect width="700" height="36" rx="8" fill="rgba(0,0,0,0.2)"/>
+    <rect y="28" width="700" height="8" fill="rgba(0,0,0,0.2)"/>
+    <text x="24" y="24" class="box-title">STEP 3 — PLAN</text>
+    <rect x="24" y="48" width="260" height="26" rx="4" fill="rgba(255,255,255,0.1)"/>
+    <text x="34" y="66" class="box-text" style="font-family: 'JetBrains Mono', monospace; font-size: 11px;">/rpi:plan "{feature-slug}"</text>
+    <rect x="24" y="82" width="500" height="24" rx="4" fill="#f3f4f6"/>
+    <text x="34" y="98" class="output-text">→ Output: rpi/{feature-slug}/plan/ [pm.md, ux.md, eng.md, PLAN.md]</text>
+    <text x="310" y="66" class="agent-text">Agents: senior-software-engineer, product-manager, ux-designer</text>
+  </g>
+
+  <!-- Arrow 4: between 689 and 704 -->
+  <g transform="translate(450, 692)">
+    <line x1="0" y1="0" x2="0" y2="10" stroke="#d1d5db" stroke-width="2"/>
+    <polygon points="0,16 -5,8 5,8" fill="#d1d5db"/>
+  </g>
+
+  <!-- Step 4: y=712, height=170, ends at 882 -->
+  <g class="clickable" transform="translate(100, 712)">
+    <rect width="700" height="170" rx="8" fill="url(#step4Grad)" filter="url(#shadow)"/>
+    <rect width="700" height="36" rx="8" fill="rgba(0,0,0,0.2)"/>
+    <rect y="28" width="700" height="8" fill="rgba(0,0,0,0.2)"/>
+    <text x="24" y="24" class="box-title">STEP 4 — IMPLEMENT</text>
+    <rect x="530" y="8" width="150" height="20" rx="10" fill="#10b981"/>
+    <text x="605" y="22" text-anchor="middle" class="gate-text" style="fill: white;">PER-PHASE VALIDATION</text>
+    <rect x="24" y="48" width="300" height="26" rx="4" fill="rgba(255,255,255,0.1)"/>
+    <text x="34" y="66" class="box-text" style="font-family: 'JetBrains Mono', monospace; font-size: 11px;">/rpi:implement "{feature-slug}"</text>
+    <rect x="24" y="82" width="652" height="50" rx="4" fill="rgba(0,0,0,0.15)"/>
+    <text x="34" y="100" class="box-text" style="font-size: 11px; font-weight: 500;">Phase Loop:</text>
+    <text x="110" y="100" class="agent-text">Discovery → Implementation → Self-Validation → Code Review → User Validation → Docs</text>
+    <text x="34" y="120" class="agent-text">Status:  [ ] Not Started   [~] In Progress   [x] PASS   [!] Conditional   [-] Failed</text>
+    <text x="350" y="66" class="agent-text">Agents: Explore, senior-software-engineer, code-reviewer</text>
+    <rect x="24" y="140" width="350" height="22" rx="4" fill="#f3f4f6"/>
+    <text x="34" y="155" class="output-text">→ Output: rpi/{feature-slug}/implement/IMPLEMENT.md</text>
+  </g>
+
+  <!-- Arrow to final: between 882 and 897 -->
+  <g transform="translate(450, 885)">
+    <line x1="0" y1="0" x2="0" y2="10" stroke="#d1d5db" stroke-width="2"/>
+    <polygon points="0,16 -5,8 5,8" fill="#d1d5db"/>
+  </g>
+
+  <!-- Final Output: y=905, height=45, ends at 950 -->
+  <g transform="translate(100, 905)">
+    <rect width="700" height="45" rx="8" fill="#111827" filter="url(#shadow)"/>
+    <text x="350" y="28" text-anchor="middle" class="box-text" style="font-size: 13px; font-weight: 500;">✓  Feature Complete — Working Code + Documentation → Ready for PR</text>
+  </g>
+
+</svg>